Summary

  • NANOG’s 2024 annual report counted 125, 184 and 209 first-time attendees at NANOG 90, 91 and 92. It also described mentoring, newcomer presentations, an ambassador programme and an event help desk. These are meaningful records of entry and activity, but the reviewed public record does not connect the counted cohorts to later participation.
  • The published metrics changed as meetings moved from physical events to virtual and hybrid formats. Registration, in-person attendance, virtual registration, peak viewership, membership status and demographic tables do not automatically share a denominator, even when they appear next to one another.
  • A missing longitudinal link is not proof of poor retention, exclusion or programme failure. Some people attend once by choice, later participation is not a moral promotion, and joining records across systems can create serious privacy, security and re-identification risks.
  • A proportionate remedy is a minimal aggregate transition ledger: stable definitions, explicit denominators, separate return and later-role counts, small-cell suppression, limited retention and no public individual trajectories.

Two useful pages and the space between them

NANOG’s 2024 annual report offers two adjacent ways of seeing participation. The first is numerical. It reports 125 first-NANOG attendees at NANOG 90, 184 at NANOG 91 and 209 at NANOG 92. The second is programmatic. Elsewhere, the report describes the Mentorship Committee’s primary focus as improving early NANOG experiences and supporting continuing professional mentorship and growth. It says the committee met semi-monthly to discuss mentee-mentor pairings, developed presentations for the three annual meetings, launched an ambassador programme in October and maintained a help desk at events.

Both views are informative. The first establishes that each meeting included a substantial number of people labelled as attending NANOG for the first time. The second establishes that NANOG operated activities intended to support newcomers and continuing professional growth. Neither view tells a reader what proportion of those 125, 184 or 209 people later returned, requested a mentor, completed a pairing, submitted a talk, volunteered, joined as a member or entered a formal role.

That absence must be described precisely. It does not mean those later events did not occur. It does not mean NANOG failed to retain newcomers, that mentoring was ineffective or that first-time attendance had no professional value. It means the reviewed public record stops before a longitudinal question can be answered. The reports count entry at a meeting and describe opportunities around entry. They do not publish an aggregate transition from the entry cohort to a later state.

The distinction matters because activity and continuity answer different questions. A newcomer orientation can be offered and well run without producing a public return-rate measure. A mentoring committee can arrange support without publishing a cohort study. A meeting can welcome hundreds of first-time attendees while making no claim that they will all return. An annual report can be accurate about each of these facts while remaining unable to show how they connect.

This is not a complaint that NANOG publishes no participation data. The opposite is closer to the truth. Across the reviewed reports, NANOG provides meeting totals, virtual and in-person figures, first-time counts or percentages, member and non-member registrations, student registrations, speakers, women-in-attendance figures, employment bands and organisation types. The record grew more granular over time. Its limitation is not the absence of snapshots; it is that snapshots are arranged by meeting rather than by cohort.

A meeting snapshot asks: how many people registered or attended this event, and what characteristics were reported about them? A cohort record asks: among people defined as entering for the first time at one event, how many reached a separately defined later state within a stated period? Those questions require different denominators, different time horizons and different privacy decisions. One cannot be answered by adding more columns to the other unless the joins and definitions are designed deliberately.

The argument of this article is therefore narrow. NANOG has made the front door increasingly visible. The reviewed public record does not make the path beyond that door measurable. A minimal, privacy-preserving transition ledger could do so without naming individuals, publishing personal histories or treating one-time attendees as failed members of a conversion campaign.

A snapshot can be accurate and still answer the wrong question

Annual reports are naturally organised around a reporting year. Meeting pages are naturally organised around an event. That structure is useful for operations: it shows the size of a meeting, its mode, selected registration characteristics and programme activity. It also makes it easy to mistake temporal proximity for a measured relationship.

Suppose one page reports a large first-time cohort and another page describes a mentorship programme. The reader can reasonably say that NANOG counted newcomers and offered newcomer support in the same period. The reader cannot infer that the counted people used the support, that those who used it returned, or that any later outcome was caused by the programme. To make even the first connection, the record would need an exposure count: how many people in the defined first-time cohort attended orientation, requested a mentor, received a match or used the help desk.

To make a later connection, it would need a time window and an aggregate transition count.

The absence of those links also prevents negative conclusions. A reader cannot take the next annual report’s meeting total, compare it with the earlier first-time count and estimate loss. Later meetings include different people, different modes and different circumstances. The original newcomers are not identified within the later aggregate. Some may return virtually, some in person, some after more than one meeting, and some not at all. The reviewed public record does not publish those paths.

Nor is “return” self-explanatory. Does a virtual registration count? Must a person collect a badge, watch a session or appear in a check-in record? Is a return at six months different from a return after 18 months? If a meeting rotates location or format, does the same window remain fair? Before a rate can be meaningful, the organisation must define the state being counted.

The same applies to later participation. A talk submission is not the same as an accepted talk. Expressing interest in volunteering is not the same as completing a volunteer task. Membership entry is not equivalent to committee service, candidacy or election. These states can overlap, and none is inherently a higher form of human value. A useful record would report each transition separately rather than compressing them into one broad measure of engagement.

This distinction protects people from being interpreted as missing data points in an institutional success story. A person may attend once for a specific technical subject and receive exactly the value sought. Another may return repeatedly without becoming a member. A third may submit a talk without volunteering. A fourth may join a committee years later. Continuity can take several forms, and non-continuity is not automatically a negative outcome.

What NANOG could learn from aggregate transitions is operational rather than moral. It could see whether first-time cohorts return under a stable definition, whether people exposed to a programme are visible in later aggregate states, and whether definitions remain comparable when meeting formats change. The ledger would not declare that any individual made the right choice. It would let the organisation test whether its public account of participation reaches beyond the event at which entry was counted.

2019: percentages that describe overlapping populations

The 2019 annual report provides a useful early example of why denominator discipline is necessary. It gives total attendance of 1,304 for NANOG 75, 971 for NANOG 76 and 1,164 for NANOG 77. On the same page, it gives first-time percentages of 26%, 28% and 24%. It separately gives member percentages of 29%, 28% and 26%.

Those first-time and member percentages describe different characteristics. They are not two parts of a single partition. A person can be attending NANOG for the first time and also hold membership, depending on the applicable rules and timing; a returning attendee can be a member or not. Adding the two percentages or treating one as the complement of the other would create an answer the report does not claim to provide.

This is a basic point, but it carries much of the later analysis. A table can place measures side by side because they illuminate the same meeting without implying that they share categories, timing or collection methods. The heading and layout may be common while the denominator underneath each percentage differs. A careful reader must resist using whichever nearby total produces a plausible result.

The same 2019 page separately displays counts for men, women, undeclared and other. The sums of those displayed counts do not equal the displayed total meeting-attendance figures. That arithmetic gap is real as a feature of the presentation. Its cause is not established in the reviewed page. The gender counts could use a responding population, a registration subset, an in-person group, a point-in-time extraction or another definition. The report does not say.

The correct conclusion is not that the annual report contains false data. It is that the denominator and collection method for the displayed gender series are not defined on the reviewed page. An explanatory note could resolve the issue by identifying who was asked, when the information was collected, how non-response was handled and which meeting population formed the base.

The label “women in attendance” also has a limited evidentiary meaning. It is the source’s category for a displayed count. It does not establish how gender identity was collected, whether the count represents all attendees or what comparison population would be appropriate for an inclusion claim. It cannot be used, without more, to conclude that a programme changed representation or that one group was selected, retained or excluded.

The 2019 record is nevertheless substantive. It gives meeting size and makes first-time presence visible as a proportion. That is more informative than a total alone. The lesson is not to discard the percentages. It is to publish the denominator definition that allows each percentage to answer one clear question, then keep that meeting-level measure separate from any later cohort transition.

For longitudinal analysis, the 2019 first-time percentages could in principle define an entry population if the underlying status were stable and recoverable. The public report does not supply the later joins needed to follow that population. It also does not establish whether “first time” was self-reported, calculated from prior registrations or derived another way. Retrospective reconstruction would therefore risk attaching modern definitions to historical figures that were collected differently.

A prospective ledger avoids that problem. It can state in advance what “first NANOG” means, which modes count and what event marks attendance. Historical tables can remain useful snapshots without being forced into a continuity series they were not designed to support.

2021: when registration and attendance changed shape

The move through virtual and hybrid meetings made the denominator problem more visible. The 2021 annual report describes NANOG 81 and NANOG 82 as virtual meetings. It reports 921 and 758 virtual registrations, alongside peak-viewership figures of 186 and 128.

Registration and peak viewership are not interchangeable. A registration records one kind of entry into the event system. Peak viewership records the greatest simultaneous audience observed under the relevant viewing measure. One person may register and never view; another may view at a different time; several viewers may share a connection; and peak concurrency does not count everyone who watched across the event. The reviewed report gives both figures, but it does not invite their merger into a single attendance count.

NANOG 83 is described as the first hybrid meeting. The report lists 300 virtual registrations and 312 in-person registrations. It also lists 214 member registrations, two student registrations, 36 speakers and 63 women in attendance, labelled 10%.

Again, proximity does not settle denominator. The page does not state which base produces the 10% label attached to the 63 women-in-attendance count. Dividing 63 by one of the adjacent registration totals because the result looks close would replace a missing definition with guesswork. The member, student and speaker counts also represent different, potentially overlapping characteristics. A speaker may be a member; a student may be attending virtually or in person; the displayed categories are not necessarily a partition.

Hybrid participation complicates first-time status as well. A person’s first virtual registration may precede a first in-person meeting. If both are counted simply as “first NANOG,” later analysis may mix different entry experiences. If only in-person attendance counts, the record needs to say so. Neither rule is inherently correct for every purpose. The choice should follow the question.

For an event-access question, a virtual first encounter may be meaningful. For an orientation designed around an in-person venue, first physical attendance may be the relevant cohort. For professional continuity, the ledger may need both: first participation in any mode and first in-person participation. The public reports reviewed here do not establish that such a distinction was applied.

This is why year-to-year comparisons can become fragile even as the tables grow more detailed. A percentage can be calculated consistently within one meeting yet lose comparability when the meaning of registration, attendance or mode changes. Stable measurement does not require pretending virtual and in-person experiences are identical. It requires naming the difference and keeping each denominator attached to its definition.

The 2021 report also shows why a continuity record cannot simply be reconstructed from public totals. A virtual registration in one year, a hybrid in-person registration later and a committee role in another system may belong to the same person, but the public tables do not identify that relationship—and should not identify it at the individual level. The institutional question is whether a temporary, purpose-limited join could produce aggregate transitions while keeping the identity link private and deleting it after the reporting window.

2022: more fields, more reasons to preserve definitions

The 2022 annual report provides a richer meeting-by-meeting view for NANOG 84, 85 and 86. It includes total, virtual and in-person figures, then member, non-member, student, first-NANOG and women-in-attendance figures. The additional detail improves the picture of each event. It also makes arithmetic gaps visible.

For NANOG 84, the report gives 777 total attendees, 309 virtual and 466 in person. It lists 156 member, 306 non-member and four student registrations. Those three status counts sum to 466, matching the in-person figure. It also lists 153 first-NANOG attendees and 93 women in attendance, labelled 11.97%.

The exact match between the three status categories and the in-person figure is informative for that meeting. It suggests those categories may have been presented as an in-person partition there. It does not prove the same relationship governs every other field or meeting. The total figure is not simply explained by that one match, and the women-in-attendance percentage still needs its own stated denominator.

For NANOG 85, the report gives 717 total attendees, 203 virtual and 514 in person. It lists 167 member, 335 non-member and 11 student registrations. Those status figures sum to 513—one fewer than the displayed in-person count. The report lists 192 first-NANOG attendees and 73 women in attendance, labelled 10%.

For NANOG 86, the report gives 1,083 total attendees, 153 virtual and 930 in person. It lists 267 member, 634 non-member and 11 student registrations. Those categories sum to 912, which is 18 fewer than the displayed in-person figure. It lists 296 first-NANOG attendees and 133 women in attendance, labelled 12.28%.

The one-person and 18-person gaps do not establish registration errors. They may reflect missing responses, another category, timing differences, changes in status after extraction or a definition not shown on the page. The reviewed public record does not say. An analyst should report the displayed numbers and the unexplained difference, then stop before assigning a cause.

These examples show why “the denominator” is often too singular. A report can have a meeting total, an in-person total, a virtual total, a status-response total and a demographic-response total. Each may be valid for its own purpose. The accountability problem arises when the reader cannot tell which base belongs to which percentage or why the base changes.

For a transition ledger, the first-NANOG count needs an especially clear base. NANOG 84’s 153, NANOG 85’s 192 and NANOG 86’s 296 are meaningful entry counts as displayed. But the reviewed report does not establish whether they include virtual registrants, in-person attendees, completed registrations, checked-in entities or another group. Without that definition, comparing the share across meetings risks confusing changes in measurement with changes in participation.

The richer field set also illustrates why later-role data cannot be inferred from status. A non-member registration is not evidence that the person never became a member. A member registration does not show when membership began or whether attendance contributed to that choice. A student registration does not describe later speaking, volunteering or employment. Snapshot labels capture a state at one moment; continuity requires a dated transition.

2023: demographic detail and the beginning of a mentoring path

The 2023 annual report continues the granular meeting record. It reports 874 total attendees at NANOG 87, with 164 virtual and 710 in person; 966 at NANOG 88, with 94 virtual and 872 in person; and 840 at NANOG 89, with 102 virtual and 738 in person. It lists 202, 209 and 176 first-NANOG attendees for the three meetings.

A separate registration table displays 26%, 22% and 20% next to those first-time counts. The report does not explicitly define the denominator for each percentage. It may be tempting to test the counts against several nearby totals and choose the one that seems to reproduce the displayed percentage. That exercise would not establish what NANOG actually used, especially where values may be rounded or tables extracted at different times.

The appropriate record is therefore twofold: the report displays the counts and the percentages, and the reviewed pages do not state the denominator definition that joins them. This is a presentation limitation, not a finding that the values are wrong. A short denominator note would make the series much easier to compare.

The 2023 report also publishes employment bands and organisation-type demographics. Those tables have their own displayed totals. They cannot be assumed to use the same base as every registration measure. Someone may decline an employment question, fit an “other” category, update a profile after registration or be counted at a different stage. The report’s value lies in showing dimensions of the meeting population. Its limitation is that the dimensions do not automatically form one fully cross-tabulated cohort.

The same year introduces a clearer programme path. The report calls 2023 the inaugural year of the DEI Committee and says it held events focused on diversity, equity and inclusion for NANOG 88 and 89. It describes the Mentorship Committee as making recommendations to the Board for consideration and final approval. It says the committee outlined a method for requesting and assigning mentors and started building a programme to support new attendees.

Those statements establish activity, purpose and part of the control structure. They do not publish the number of mentor requests, assignments, completed pairings or later return visits. They do not establish whether attendees exposed to the DEI events later entered any role. They do not show programme success or failure.

That boundary is particularly important because programme language can sound longitudinal even when reporting remains annual. “Supporting new attendees” expresses an intended population. A method for requesting and assigning mentors describes an operating step. Neither is an outcome measure. The missing bridge could be modest: among a defined first-time cohort, how many were offered orientation, requested mentoring, received a match and completed the agreed contact period? At later intervals, how many returned or entered another separately defined state?

Even that design would not establish causation by itself. People who request mentors may differ from people who do not. A return could reflect employer travel, technical interest, location, personal relationships or many other conditions. The ledger could show observed transitions and programme exposure. A credible causal evaluation would require a different and much more demanding design, which the evidence and the proportionate remedy here do not call for.

2024: entry counts become clearer, continuity remains open

The 2024 report brings the opening question into sharp focus. NANOG 90 had 734 total attendees, 105 virtual and 629 in person, with 125 first-NANOG attendees. NANOG 91 had 738 total, 65 virtual and 673 in person, with 184 first-NANOG attendees. NANOG 92 had 869 total, 61 virtual and 808 in person, with 209 first-NANOG attendees.

The registration table gives first-time percentages of 20%, 27% and 26%. Those rounded percentages align arithmetically with the displayed in-person counts: the first-time figures are close to the stated shares of 629, 673 and 808. That is useful evidence about the likely base. The table still does not separately define the denominator in words, so the relationship should be described as an arithmetic alignment rather than a definitive semantic rule.

The report also provides member, non-member, student, women-in-attendance, employment and organisation-type figures. This is a substantial event record. It can help NANOG understand meeting composition and communicate the scale of newcomer entry. It still does not show whether the people counted at NANOG 90 are included among later attendees at NANOG 91 or 92.

The Mentorship Committee account makes the gap more consequential because it states an intention that extends beyond initial arrival. The primary focus is described as improving early NANOG experiences and supporting continuing professional mentorship and growth. The committee met semi-monthly to discuss pairings, developed presentations for all three annual meetings, launched an ambassador programme in October and operated a help desk. Recruitment is listed among four growth pillars.

NANOG describes a “successful 2024.” That assessment should be attributed to the organisation. The reviewed report does not publish a entity outcome definition against which an outside reader could test the description. It does not give the denominator of potential mentees, the number who requested a mentor, the number matched, the number completing a defined pairing or the number later returning.

This does not invalidate NANOG’s assessment. A committee may judge success through delivery, feedback, quality of interactions or achievements not reduced to a public table. The public record simply does not tell the reader which measure supports the assessment. A transition ledger would allow the organisation to preserve qualitative judgement while adding a limited set of observable operating outcomes.

The distinction between delivery and effect is useful here. Newcomer presentations can be developed for all three annual meetings. An ambassador programme can launch. A help desk can be staffed. Those are outputs under organisational control. Return, speaking or committee service occur later and depend on entity choice as well as opportunity. A good record would not hold the committee solely responsible for every later state. It would show the transition while acknowledging that the mechanism is not established.

2026: NANOG names the paths it invites

The NANOG 97 agenda scheduled a Newcomers Orientation. That establishes that an orientation was offered within the public programme. It does not establish how many people attended, whether they were all first-time attendees or what happened afterward.

A June 2026 message in NANOG’s attendee archive adds a post-meeting view. It reported that 147 survey responses had been received across three days and expressed an aspiration to reach 200. The 147 is an interim response count. It is not the meeting attendance total, not a response rate and not evidence that respondents represent the full meeting population. The reviewed message does not provide the invitation denominator or respondent composition.

The message also invited people to remain involved in several ways: volunteering, joining a committee, nominating someone for the Board, participating in elections, donating, submitting a talk and welcoming a newcomer. These invitations are analytically valuable because they identify later participation states in NANOG’s own public communication. They show that the organisation offers more than one path after a meeting.

The invitations do not establish that a first-time attendee has progressed along any of those paths. They also do not make the paths equivalent. Submitting a talk is different from speaking; nominating someone is different from standing as a candidate; participating in an election is different from holding membership if membership eligibility governs the act; welcoming a newcomer can be informal and need not create a formal record.

Still, the message suggests a practical transition map grounded in the organisation’s own activities. Entry can be followed by return, programme participation, speaking interest, volunteer service, membership or a formal role. The public measurement gap is that the reviewed record does not publish how a first-time cohort intersects with any of these later states.

Survey reporting deserves its own branch. A transition ledger should record how many people were invited, how many responded and the resulting response rate, while making non-response explicit. Survey answers should not be treated as representative unless the design supports that claim. Nor should completion become a condition of being counted as engaged. A person can participate meaningfully and decline a survey.

The 2026 communication therefore reinforces both sides of the argument. NANOG visibly invites continuity. It also demonstrates why simple activity counts are not enough. An orientation on the agenda, an interim survey count and a list of opportunities each answer a valid question. None supplies the denominator and time window needed to describe what a first-time cohort did later.

The strongest case against longitudinal tracking

The case for restraint is substantial. NANOG already publishes more meeting-level detail than many event organisations. Building a longitudinal record would require care, coordination and sustained ownership. The fact that a question can be asked does not mean every underlying link should be created.

First, participation is not a product funnel. People do not attend a technical meeting under an obligation to advance through institutional stages. A first-time attendee who never returns may have obtained the needed knowledge, made useful contacts or simply chosen a different venue. A returning attendee who never becomes a member is not incomplete. Membership, speaking, volunteering and leadership are distinct options, not ranks on a universal ladder.

Second, the relevant states may live in separate operational systems for legitimate reasons. Registration supports event access. Membership records support organisational rights and services. Talk submissions support programme selection. Volunteer and committee records support particular tasks. Election records support formal governance. Combining those systems can expand the purpose for which personal information is used.

Third, a linked record can become identifying even if the published output is aggregate. Small cohorts and rare transitions make re-identification possible. If only a few first-time attendees later serve in a formal role, people familiar with the community may infer who they are. Cross-tabulating meeting, employer, geography, programme exposure and later role would magnify the risk.

Fourth, retention creates its own danger. Measuring return at 18 months requires a linkage key to persist across a period. Keeping that key indefinitely would make future use easier but would violate the principle of collecting only what is needed for a defined question. Security exposure grows with the number of systems connected and the length of time identifiers remain available.

Fifth, consent is complicated. Asking people to opt into evaluation may produce a biased group, because those already inclined toward deeper participation may be more likely to agree. Making consent too prominent could also make newcomers feel observed. Yet relying on operational necessity to join records without a clear public explanation could undermine trust.

Sixth, administration is not free. Definitions must be maintained across meeting formats. Cancellations, duplicate registrations, email changes and membership timing need rules. Small cells must be reviewed before publication. Staff and volunteers may reasonably prefer to spend time improving events rather than producing elaborate longitudinal reports.

Seventh, sensitive identity collection is neither necessary nor justified by this thesis. The reviewed record does not support demanding race, gender, disability, immigration, employer or financial dossiers. Such information can be deeply personal, legally constrained and prone to misuse. A minimal continuity question can be answered without expanding collection into protected or intimate characteristics.

Finally, snapshots remain useful. Registration counts help plan meetings. Mode counts describe access. Membership status can inform services. Speaker and student counts illuminate event composition. The inability to answer a causal or longitudinal question does not make those measures defective.

These objections should shape the remedy, not merely appear as a paragraph before an expansive proposal. The ledger must be minimal, aggregate, purpose-limited and designed around deletion. It should never publish individual trajectories. It should not treat a later role as proof of personal success or an absent transition as failure. Where a lawful, technical or re-identification constraint makes a field unsafe, the public record should say so and omit it.

The minimum transition ledger

The smallest useful ledger begins with definitions. For every meeting, NANOG could state what it means by registered, attended, virtual, in person and first NANOG. It should explain whether first-time status is self-reported, calculated from prior records or derived another way. It should say how cancellations, no-shows, duplicate registrations and mixed-mode participation are treated.

The first row would give the entry cohort by modality. It might distinguish first participation in any mode from first in-person participation if both matter to programme design. The denominator for every percentage would be written next to the measure, not left to arithmetic reconstruction.

The second row would report aggregate return at fixed windows—six, 12 and 18 months are reasonable operating intervals to consider. Each window should define what counts as return. The ledger should not label non-return as loss. It should state only how many people in the original cohort met the return definition within the period.

The third row would cover programme exposure. For orientation, it could report invitations and attendance if attendance is measured appropriately. For mentoring, it could report requests, matches and completion under a modest, pre-stated definition. For ambassador activity or a help desk, it should count exposure only if doing so is feasible without intrusive observation. These are operating measures, not proof that the programme caused a later state.

The fourth row would report talk participation within a fixed window: submissions and speaking appearances as separate aggregates. A submission demonstrates one kind of active interest; acceptance depends on programme selection and available slots. Combining the two would obscure the difference between entity action and institutional decision.

The fifth row would report volunteer interest and completed service separately. Clicking an interest form or writing to a committee is not the same as performing a task. A completion definition could be simple and appropriate to the role. Informal contributions should not be forced into surveillance merely to improve the count.

The sixth row would report membership entry within the window, while making clear that membership is not superior to other forms of participation. The relevant date should be defined, and people already holding membership at first attendance should not be counted as later entries. This is precisely why first-time and member characteristics cannot be treated as mutually exclusive.

The seventh row would cover formal roles—committee service, candidacy or another defined position—only when the cell is large enough to publish safely. These states may be rare, so suppression may be the norm. The ledger should never list the individuals who form a transition cell or allow several tables to be combined into an obvious identity.

The eighth row would treat surveys honestly. It should publish invitations, responses and response rate. If a survey is sent only to a subset, that subset should be defined. Results should carry a warning about non-response and should not be treated as the voice of the whole meeting without supporting evidence.

Each row needs a status vocabulary. A value of zero means the transition was measured and no one met the definition. “Not collected” means the measure was outside the design. “Unavailable” means the organisation cannot establish the value. “Suppressed” means a known value is withheld under the small-cell rule. Those states should never be collapsed into a blank.

Privacy controls belong in the design. Public tables would contain only aggregates. A temporary linkage key could connect the entry cohort to later systems for the defined window, with access restricted to the people responsible for producing the aggregate record. The key should be destroyed after the final window and validation period. The published totals can remain; the individual join should not.

The ledger should also publish a minimum-cell threshold and guard against combination risk. A cell that appears safe alone may become identifying when paired with meeting, role or programme data. Where risk remains, NANOG could publish a broader category, combine time periods or state that the value is suppressed.

No protected-characteristic expansion is required. If NANOG already collects demographic information for a separate, legitimate purpose, any publication decision should follow its own consent and privacy rules. The continuity ledger proposed here can begin with non-sensitive transition states and need not create personal profiles.

Ownership across organisational boundaries

A ledger that joins several systems will fail if no durable function owns the definitions. The reviewed public record describes multiple control points. Staff and registration systems collect meeting-level information. Event and communications teams operate orientation, surveys and invitations. The Mentorship Committee develops methods and supports pairings, while the Board considers and approves programme recommendations. Membership, programme selection, committees and elections govern later participation states.

The reviewed sources do not identify one owner for the joins between those areas. That absence is again scoped to the public record; it does not prove that no one coordinates the information within NANOG. Publicly naming a steward would nevertheless improve accountability. The steward’s role would be narrow: maintain definitions, confirm that each denominator matches its question, coordinate aggregate extraction, apply suppression and ensure deletion on schedule.

The steward should not be authorised to build broad personal profiles. Nor should a mentorship body be judged solely on later membership or leadership counts it cannot control. Responsibility must follow the transition. Registration staff can define entry. Programme teams can define exposure and selection. Membership and committee functions can define their later states. The steward can ensure the published joins use comparable windows and do not erase the differences.

Board oversight is appropriate for the purpose and boundary of the record, especially because the Board approves organisational programmes and committee structures. Operational calculation can remain with the relevant functions. This separation prevents a single committee from becoming both programme operator and sole evaluator while avoiding a new bureaucracy.

Changes in meeting mode require an explicit review of definitions. A shift from in-person to virtual, or to another hybrid form, should not silently alter the entry cohort. The public note can explain whether the series continues, splits by mode or begins a new baseline.

Changes in programmes require similar care. If orientation, mentoring or ambassador activity changes, the exposure measure should change visibly. Historical figures should not be restated under a later definition. Stable reporting does not mean preventing innovation; it means preserving enough version history to understand what changed.

Implementation without rebuilding history

The safest starting point is prospective. NANOG could select a future meeting, publish definitions before registration closes and create the entry cohort under those rules. It could then publish initial meeting figures, a six-month return update, a 12-month participation update and a final 18-month aggregate record.

Retrospective analysis of 2019–2024 should remain limited to what the reports already establish. The 2019 percentages, 2021 virtual and hybrid measures, 2022 field counts, 2023 first-time series and 2024 entry cohorts should not be forced into a single longitudinal baseline. Their value is to show the evolution of reporting and the need for stable definitions.

Before any linkage begins, NANOG should write a purpose statement. The purpose would be to understand aggregate continuity from first participation to selected later states and to improve programme learning. It would explicitly exclude public individual tracking, ranking attendees, assessing personal worth, compiling sensitive identity histories or using non-return as a negative label.

Next comes a data-minimisation review. For each proposed transition, the organisation should ask whether the state is necessary, whether it can be measured reliably, how long a join must exist and whether a safer aggregate can answer the question. If a field cannot survive that review, it should not be collected for the ledger.

The measurement schedule should be modest. Meeting-level entry figures can publish with the annual report or event record. Six-, 12- and 18-month updates can follow on a predictable calendar. After the final update and validation, linkage keys should be destroyed according to the declared retention schedule.

Quality checks should focus on definitions and arithmetic rather than an institutional performance rating. Does the denominator match the label? Are mode changes visible? Do status categories account for missing responses? Are suppressed cells distinct from zero? Can two published tables be combined to identify someone? Those questions protect both analytical usefulness and entity privacy.

NANOG should also publish what the ledger cannot establish. Observed association is not programme causation. A return is not necessarily caused by mentoring. A talk submission does not measure belonging. Membership entry does not prove greater community value. A formal role does not convert attendance into a representative mandate.

What would weaken the critique

The argument is falsifiable. It would weaken materially if NANOG publishes stable cohort definitions and aggregate transitions from first attendance to return and later participation states, using explicit denominators, comparable windows, small-cell suppression and limited retention.

It would also narrow if NANOG publishes a specific legal, consent, technical or re-identification constraint that makes a proposed transition disproportionate. Not every desirable question has a safe answer. A clear reason for omitting a field is evidence of considered data governance, not an accountability failure.

The critique would need revision if the reviewed gaps are already addressed in other public material. This article is limited to the eight-reference listed below. It does not claim that NANOG lacks private operating measures or that no other public page exists. New evidence should change the analysis.

Conversely, more meeting-level columns would not by themselves resolve the thesis. A richer snapshot remains a snapshot unless it connects a defined entry cohort to a later aggregate state. Programme activity counts would improve operating visibility but would not alone establish continuity or effect.

Conclusion: count the transition, not the person

NANOG’s public participation record is substantial. The annual reports show meeting scale across physical, virtual and hybrid formats. They identify first-time cohorts, membership status, students, speakers, women in attendance, employment bands and organisation types. The programme record describes orientation, mentoring, ambassadors, a help desk, surveys and invitations to volunteer, submit talks or take part in formal governance.

These disclosures make entry and available paths visible. The reviewed public record does not publish the aggregate bridge between them. It does not show what proportion of a first-time cohort returns within a defined period, enters a programme, submits a talk, volunteers, becomes a member or serves in a formal role.

That gap should not be inflated into a story of failure. A one-time attendee may have achieved the intended purpose. A programme can deliver real value without causing a later institutional role. Separate systems and strong privacy boundaries can be legitimate. Small cohorts can make even aggregate reporting unsafe.

The proportionate response is not a public list of journeys. It is a small set of defined transitions, each with its own denominator, time window and suppression rule. The organisation can retain aggregate history while destroying the temporary link that produced it. It can report uncertainty without turning blanks into zeros and activity without turning activity into causation.

The 125, 184 and 209 first-time attendees counted in 2024 are not evidence of retention. They are not evidence of loss either. They are entry cohorts whose later aggregate states are not published in the reviewed record. NANOG already shows the door and names several paths beyond it. A minimal transition ledger would show whether those paths are used, while leaving the people who use—or decline—them out of public view.

Sources